Host media processing
Encyclopedia
Host Media Processing is a term used to refer to a telephony
system that uses the millions of instructions per second
(MIPS) of processing power on a general-purpose computer to process a telephony call’s media stream rather than use digital signal processor
s (DSPs) to perform the task. When telephony call streams started to be digitized for time-division-multiplexed (TDM
) transport, processing of the media stream, to enhance it in some way, became common. For example, digital echo cancellers were added to long-haul circuits, and transport channels were shaped to improve modem performance. Then, in the mid-‘80s, computer-based systems that implemented messaging, for example, used DSPs to compress the audio for storage, and fax servers used DSPs to implement fax modems.
However, since the late ‘90s, the MIPS available on low-cost PCs
have been adequate to process several media streams, while still leaving enough processing power to handle the application. And, following Moore’s Law, PC capacity continues to double every 18 months, while the MIPS required to process a call’s media stream have remained relatively constant. Now, in the latter half of the century’s first decade, a single PC can handle well over 100 simultaneous calls.
Prior to IP telephony, when you wanted to connect a telecommunications system to a telecom network it was necessary to have a telecom-specific physical interface. This could mean an analog interface (POTS/DS-0), for low-density non-network systems, or a digital interface, such as a T-1
or E-1 line (DS-1
, delivering 24 or 32 DS-0s). A DS-4 connection delivers 274.176 Mbit/s or 4032 DS-Os. In each case, telecom-specific electronic interfaces, which were proprietary and, therefore, relatively expensive, were necessary. The situation changes dramatically with an all-IP telecom infrastructure. The network interfaces move from being a significant proprietary component to off-the-shelf high-performance IP interfaces, an inherent feature in every modern computing system. Today, 10-Gigabit
Ethernet
' telephony systems are being deployed.
The term Host Media Processing was first used in a product name by Intel in the early 2000's. It was quickly adopted as a generic term for software-based telephony products, used by many companies including Aculab, Pika, Eicon Networks, Uniqall, Commetrex, and NMS. Intel's Host Media Processing product line (still called HMP) exists today under the Dialogic
banner.
The concept of using an industry standard PC to do telephony processing is now widely understood and accepted, with open-source platforms like Asterisk
, YATE and FreeSWITCH
using the same principle. The rise of interest in VoIP and Fax-over-IP (FoIP) have driven demand for open, host-based solutions that can be molded into a variety of different communications solutions. HMP components are used today to implement many different kinds of solutions including PBX, conference servers, unified communications servers and IVR
. The emergence of virtualization
in recent years also increases the appeal of HMP, since it is then possible to think of telephony resources as being virtual channels (rather than dedicated hardware boards), which offer the same benefit as virtual processors and servers, i.e. resilience; less hardware; space saving; lower maintenance.
Network connectivity through low-cost industry-standard interfaces influences the consideration of whether to use DSPs or server blades for media processing, especially in media servers, where packet-delays are not as troublesome and TDM interfaces are not required. Without telephony interface blades and their attendant chassis and power systems available to host the DSPs, the addition of DSPs on proprietary blades must be independently justified. They will continue to be justified for the highest-density applications. However, with the semiconductor industry continuing to follow Moore’s law, host media processing will support 1500 channels on one blade in 2010. DSPs will always offer even higher densities, but if 1500 channels meets the system requirement, higher densities will have little incremental value.
Not every use of the term “HMP” means the same thing. There are, for example, HMP systems that do no actual media processing, so it is important to understand how the term is being used today.
Modern digital-media telephony systems require signal processing to transform a call stream or extract information from it. Transformation includes the processing required to send or receive a fax and to transcode the stream from one speech codec to another for capability matching or bandwidth reduction. DTMF detection, caller ID, and in-band call-progress analysis
are good examples of information extraction.
There are many limited-function media servers on the market that don’t actually do any media (signal) processing. There is an Internet Engineering Task Force (IETF) “RFC” (2833) that defines how a gateway can perform the in-band-tone analysis to extract some of the embedded information, such as DTMF and caller ID. In this case, all the media server need do is parse the RTP buffers from a gateway to derive the tone information.
But what about transcoding, where one voice-compression scheme (vocoder) it transcoded to another? Some media servers, for example, simply process buffers, and, therefore, cannot perform any transcoding, limiting them to low-function voice messaging. RTP packets are simply stored and played back as they are received. This means no AGC
, volume control, time-scale modification (playback speedup and slowdown), or capabilities matching with endpoint terminals, making this type of so-called HMP media server a viable option only in the most functionally constrained applications.
For years, the terms “signal processing
” and “media processing” have been used interchangeably, so, most appropriately, the term HMP is reserved for those systems where host MIPS are actually used to perform digital signal-processing tasks.
Telephony
In telecommunications, telephony encompasses the general use of equipment to provide communication over distances, specifically by connecting telephones to each other....
system that uses the millions of instructions per second
Instructions per second
Instructions per second is a measure of a computer's processor speed. Many reported IPS values have represented "peak" execution rates on artificial instruction sequences with few branches, whereas realistic workloads typically lead to significantly lower IPS values...
(MIPS) of processing power on a general-purpose computer to process a telephony call’s media stream rather than use digital signal processor
Digital signal processor
A digital signal processor is a specialized microprocessor with an architecture optimized for the fast operational needs of digital signal processing.-Typical characteristics:...
s (DSPs) to perform the task. When telephony call streams started to be digitized for time-division-multiplexed (TDM
Time-division multiplexing
Time-division multiplexing is a type of digital multiplexing in which two or more bit streams or signals are transferred apparently simultaneously as sub-channels in one communication channel, but are physically taking turns on the channel. The time domain is divided into several recurrent...
) transport, processing of the media stream, to enhance it in some way, became common. For example, digital echo cancellers were added to long-haul circuits, and transport channels were shaped to improve modem performance. Then, in the mid-‘80s, computer-based systems that implemented messaging, for example, used DSPs to compress the audio for storage, and fax servers used DSPs to implement fax modems.
However, since the late ‘90s, the MIPS available on low-cost PCs
Personal computer
A personal computer is any general-purpose computer whose size, capabilities, and original sales price make it useful for individuals, and which is intended to be operated directly by an end-user with no intervening computer operator...
have been adequate to process several media streams, while still leaving enough processing power to handle the application. And, following Moore’s Law, PC capacity continues to double every 18 months, while the MIPS required to process a call’s media stream have remained relatively constant. Now, in the latter half of the century’s first decade, a single PC can handle well over 100 simultaneous calls.
Prior to IP telephony, when you wanted to connect a telecommunications system to a telecom network it was necessary to have a telecom-specific physical interface. This could mean an analog interface (POTS/DS-0), for low-density non-network systems, or a digital interface, such as a T-1
Digital Signal 1
Digital signal 1 is a T-carrier signaling scheme devised by Bell Labs. DS1 is a widely used standard in telecommunications in North America and Japan to transmit voice and data between devices. E1 is used in place of T1 outside North America, Japan, and South Korea...
or E-1 line (DS-1
Digital Signal 1
Digital signal 1 is a T-carrier signaling scheme devised by Bell Labs. DS1 is a widely used standard in telecommunications in North America and Japan to transmit voice and data between devices. E1 is used in place of T1 outside North America, Japan, and South Korea...
, delivering 24 or 32 DS-0s). A DS-4 connection delivers 274.176 Mbit/s or 4032 DS-Os. In each case, telecom-specific electronic interfaces, which were proprietary and, therefore, relatively expensive, were necessary. The situation changes dramatically with an all-IP telecom infrastructure. The network interfaces move from being a significant proprietary component to off-the-shelf high-performance IP interfaces, an inherent feature in every modern computing system. Today, 10-Gigabit
Gigabit
The gigabit is a multiple of the unit bit for digital information or computer storage. The prefix giga is defined in the International System of Units as a multiplier of 109 , and therefore...
Ethernet
Ethernet
Ethernet is a family of computer networking technologies for local area networks commercially introduced in 1980. Standardized in IEEE 802.3, Ethernet has largely replaced competing wired LAN technologies....
' telephony systems are being deployed.
The term Host Media Processing was first used in a product name by Intel in the early 2000's. It was quickly adopted as a generic term for software-based telephony products, used by many companies including Aculab, Pika, Eicon Networks, Uniqall, Commetrex, and NMS. Intel's Host Media Processing product line (still called HMP) exists today under the Dialogic
Dialogic Corporation
Dialogic Corporation is a designer, developer and manufacturer of communication products. It was founded in 1984 in Montreal, Canada. Dialogic products are sold worldwide through a large network of distributors and resellers, and supplied to OEMs....
banner.
The concept of using an industry standard PC to do telephony processing is now widely understood and accepted, with open-source platforms like Asterisk
Asterisk (PBX)
Asterisk is a software implementation of a telephone private branch exchange ; it was created in 1999 by Mark Spencer of Digium. Like any PBX, it allows attached telephones to make calls to one another, and to connect to other telephone services including the public switched telephone network and...
, YATE and FreeSWITCH
Freeswitch
FreeSWITCH is a free and open source communications software for the creation of voice and messaging products. It is licensed under the Mozilla Public License , a free software license...
using the same principle. The rise of interest in VoIP and Fax-over-IP (FoIP) have driven demand for open, host-based solutions that can be molded into a variety of different communications solutions. HMP components are used today to implement many different kinds of solutions including PBX, conference servers, unified communications servers and IVR
Interactive voice response
Interactive voice response is a technology that allows a computer to interact with humans through the use of voice and DTMF keypad inputs....
. The emergence of virtualization
Hardware virtualization
Computer hardware virtualization is the virtualization of computers or operating systems. It hides the physical characteristics of a computing platform from users, instead showing another abstract computing platform...
in recent years also increases the appeal of HMP, since it is then possible to think of telephony resources as being virtual channels (rather than dedicated hardware boards), which offer the same benefit as virtual processors and servers, i.e. resilience; less hardware; space saving; lower maintenance.
Network connectivity through low-cost industry-standard interfaces influences the consideration of whether to use DSPs or server blades for media processing, especially in media servers, where packet-delays are not as troublesome and TDM interfaces are not required. Without telephony interface blades and their attendant chassis and power systems available to host the DSPs, the addition of DSPs on proprietary blades must be independently justified. They will continue to be justified for the highest-density applications. However, with the semiconductor industry continuing to follow Moore’s law, host media processing will support 1500 channels on one blade in 2010. DSPs will always offer even higher densities, but if 1500 channels meets the system requirement, higher densities will have little incremental value.
Not every use of the term “HMP” means the same thing. There are, for example, HMP systems that do no actual media processing, so it is important to understand how the term is being used today.
Modern digital-media telephony systems require signal processing to transform a call stream or extract information from it. Transformation includes the processing required to send or receive a fax and to transcode the stream from one speech codec to another for capability matching or bandwidth reduction. DTMF detection, caller ID, and in-band call-progress analysis
Call progress analysis
Call Progress Analysis , also called Call Progress Detection , is a generic term for signal processing algorithms that operate on audio during call setup. The goal of CPA is to determine the nature of the callee or the outcome of call setup to an external network...
are good examples of information extraction.
There are many limited-function media servers on the market that don’t actually do any media (signal) processing. There is an Internet Engineering Task Force (IETF) “RFC” (2833) that defines how a gateway can perform the in-band-tone analysis to extract some of the embedded information, such as DTMF and caller ID. In this case, all the media server need do is parse the RTP buffers from a gateway to derive the tone information.
But what about transcoding, where one voice-compression scheme (vocoder) it transcoded to another? Some media servers, for example, simply process buffers, and, therefore, cannot perform any transcoding, limiting them to low-function voice messaging. RTP packets are simply stored and played back as they are received. This means no AGC
Automatic gain control
Automatic gain control is an adaptive system found in many electronic devices. The average output signal level is fed back to adjust the gain to an appropriate level for a range of input signal levels...
, volume control, time-scale modification (playback speedup and slowdown), or capabilities matching with endpoint terminals, making this type of so-called HMP media server a viable option only in the most functionally constrained applications.
For years, the terms “signal processing
Signal processing
Signal processing is an area of systems engineering, electrical engineering and applied mathematics that deals with operations on or analysis of signals, in either discrete or continuous time...
” and “media processing” have been used interchangeably, so, most appropriately, the term HMP is reserved for those systems where host MIPS are actually used to perform digital signal-processing tasks.